Using in silico tools to augment physical experiments can help identify manufacturability issues early in development. That’s according to an AI technology company that spoke on a recent panel.
BigHat Biosciences, which was among the companies presenting at PEGS Boston, explained that developing in silico models of antibody yields using a cell-free expression system allows the exploration of a wider range of mutations.
And this, in turn, lets companies better optimize their product for manufacturability.
“There’s only so many experiments you can do by putting an antibody into CHO [Chinese Hamster Ovary] cells,” explains Hunter Elliott, PhD, vice president of machine learning at BigHat Biosciences.
“With in silico tools augmenting that exploration side, we can build models that make predictions for improved sequences, screening many more antibodies in silico than we need to send to the lab.”
Elliott’s talk came alongside a panelist who argued it was important for preclinical researchers to communicate with the manufacturing departments of their companies to make a success of the latest generation of harder-to-manufacture drugs.
Speaking about his own products, Elliott argues that using in silico tools to explore a wider range of possible antibody mutations means it’s possible to select the handful with the highest possible yields as well as other improved biophysical properties.
“You’re derisking your processes because you’re combining your experiments with the in silico tools you’re using,” he says.
Talking about the panel, Elliott adds that discussion around the limitations of models included the lack of publicly available data on manufacturability and developability, especially for potential products that have failed to make it to the clinic.
Some have expressed concern that using in silico tools might also accidentally screen out the best-performing antibodies. Although, he says, “my personal opinion is that the ability to predict properties of sequences without sending them into the lab makes it easier for us to optimize from a suboptimal starting sequence.”
“With these models, we can keep this imperfect antibody in the loop and take it forward through several rounds of optimization, and then, instead of a candidate molecule being killed early on, it might be engineered into manufacturability.”
